TWI858369B - Cardiac assessment method and computer program product thereof - Google Patents

Abstract

The present invention provides a heart assessment method, comprising the following steps: providing an electrocardiogram signal; obtaining a body surface potential vector according to the electrocardiogram signal; calculating the body surface potential vector with a pre-established mapping tensor to obtain a heart potential vector, wherein the mapping tensor is obtained by generating a plurality of abnormal positions and a plurality of abnormal degrees according to a reference heart; and classifying the abnormal position and abnormal degree of an assessed heart corresponding to the electrocardiogram signal according to the heart potential vector. In addition, the present invention also provides a computer program product that can execute the above-mentioned heart assessment method.

Description

Cardiac assessment method and computer program product thereof

The present invention provides a cardiac assessment method and a computer program product, and in particular, provides a cardiac assessment method suitable for early diagnosis and treatment tracking, and a computer program product capable of executing the aforementioned cardiac assessment method.

Cardiovascular diseases have been among the top ten causes of death in Taiwan and abroad for many years, including myocardial infarction and heart failure. For myocardial infarction, a cardiac catheterization procedure is generally used for diagnosis. A coin-sized incision is made in the subject's inguinal canal or axilla, and the catheter is inserted into the myocardial area to be tested along the vein. The blood vessels and the internal conditions of the heart are detected by contrast agents or endoscopes, or by non-invasive methods such as nuclear photography, computer tomography, and cardiac ultrasound. In addition, the electrocardiogram data of the subject at rest or during exercise can also be recorded and judged.

However, since cardiac catheterization is an invasive test, the operation takes a long time and consumes more medical resources. At the same time, patients need a longer recovery time after the operation. In addition, whether it is cardiac catheterization, nuclear radiography, computer tomography or cardiac ultrasound, it requires a dedicated testing space and equipment in the hospital, and professional medical personnel to operate and interpret the test results, which not only increases the cost of the test, but also reduces the convenience of the test.

On the other hand, although electrocardiograms can be used for routine clinical testing, existing electrocardiogram information only targets some symptoms of acute myocardial infarction and cannot provide effective information for atypical myocardial infarction or its early symptoms.

The inventor then devoted all his efforts to research and developed a cardiac assessment method that can classify the abnormal location and degree of the heart through electrocardiogram signals, in order to achieve convenience, reduce testing costs, reduce dependence on professional manpower, and be suitable for early prediction and treatment tracking.

The present invention provides a heart assessment method, comprising the following steps: providing an electrocardiogram signal; obtaining a body surface potential vector according to the electrocardiogram signal; calculating the body surface potential vector with a pre-established mapping tensor to obtain a heart potential vector, wherein the mapping tensor is obtained by generating a plurality of abnormal positions and a plurality of abnormal degrees according to a reference heart; and classifying the abnormal position and abnormal degree of an assessed heart corresponding to the electrocardiogram signal according to the heart potential vector.

In one embodiment, the reference heart is a virtual heart, and the heart assessment method further includes the following steps: establishing a model of the virtual heart; simulating different degrees of abnormalities at different locations of the model and generating a plurality of sample signals; and classifying the sample signals to generate a plurality of corresponding modal vectors, and establishing a mapping tensor based on the modal vectors.

In one embodiment, the heart assessment method further includes the following steps: determining at least one modal vector corresponding to the body surface potential vector among these modal vectors through machine learning; obtaining at least one weighted eigenvalue of the body surface potential vector on the modal vector and at least one corresponding weight vector; and combining the weighted eigenvalue and the weight vector to obtain the heart potential vector.

In one embodiment, the above-mentioned machine learning determines the modal vector through sparse representation classification, and the heart assessment method obtains the weighted eigenvalue and weight vector through coordinate descent.

In one embodiment, the heart assessment method further includes the following steps: pre-processing the electrocardiogram signal to obtain a plurality of local potential signals, wherein the number of local potential signals is the same as the number of leads of the electrocardiogram signal; and extracting at least one characteristic value from each of these local potential signals, and obtaining a body surface potential vector based on the characteristic value.

In one embodiment, the above-mentioned characteristic values include a J-point amplitude, a T-wave minimum amplitude, a T-wave maximum amplitude, a first potential point amplitude, and a second potential point amplitude of a local potential signal corresponding to these local potential signals, wherein the first potential point is a potential point at 0.25 times the J-T interval from the J point on the S-T segment of the local potential signal, and the second potential point is a potential point at 0.5 times the J-T interval from the J point on the S-T segment of the local potential signal.

In one embodiment, the reference heart is a virtual heart, and the heart assessment method further includes the following steps: creating a three-dimensional virtual heart image based on the abnormal position and degree of abnormality of the virtual heart.

In one embodiment, the heart assessment method further includes the following steps: projecting the three-dimensional virtual heart image onto a plane and filtering it to obtain a two-dimensional heart zoning image.

In one embodiment, the lead number of the above-mentioned electrocardiogram signal is less than 13.

In addition, the present invention also provides a computer program product, which can execute the steps performed by the above-mentioned heart assessment method when the above-mentioned computer program is loaded and executed by a computer.

Thus, the heart assessment method of the present invention can calculate the electrocardiogram signal with the pre-established mapping tensor to classify the abnormal location and degree of the heart, which can not only improve the convenience of assessment, but also reduce the manpower and cost of detection, and thus be suitable for early prediction and treatment tracking.

In order to make the above features and advantages of the present invention more clearly understood, the following is a detailed description of the embodiments with the accompanying drawings.

1~11: Area

P ₁ : Point J

P ₂ : T wave lowest point

P ₃ : T wave peak

P ₄ : First potential point

_P5 : Second potential point

P, Q, R, S, T, U: wave

S100, S100a: Database establishment phase

S200, S200a, S200b, S200c, S200d: Evaluation stage

S110~S140, S210~S260: Steps

Figure 1 is a schematic diagram of the steps of the first embodiment of the heart assessment method of the present invention.

Figure 2 is a schematic diagram of (a) the left anterior descending artery; (b) the left circumflex artery; (c) the anterior branch of the right coronary artery; and (d) the posterior branch of the right coronary artery in a virtual heart model.

Figure 3 is a schematic diagram of the steps of the second embodiment of the heart assessment method of the present invention.

Figure 4 is a schematic diagram of the steps of the third embodiment of the heart assessment method of the present invention.

FIG5 is a schematic diagram of extracting feature values in step S220b of FIG4.

Figure 6 is a schematic diagram of the steps of the fourth embodiment of the heart assessment method of the present invention.

Figure 7 is a schematic diagram of the steps of the fifth embodiment of the heart assessment method of the present invention.

FIG8 shows (a) a topographic map of the three-dimensional virtual heart image created in step S250 of FIG7; and (b) a colored topographic map.

Figure 9 is a schematic diagram of the step flow of the sixth embodiment of the heart assessment method of the present invention.

FIG. 10 is a schematic diagram of the two-dimensional cardiac zoning images of (a) the first subject; (b) the second subject; and (b) the third subject obtained in step S260 of FIG. 9 .

The above-mentioned and other technical contents, features and effects of the present invention will be clearly presented in the detailed description of the preferred embodiments with reference to the drawings below. It is worth mentioning that the directional terms mentioned in the following embodiments, such as up, down, left, right, front or back, etc., are only reference directions of the attached drawings. Therefore, the directional terms used are for explanation rather than limitation of the present invention. In addition, in the following embodiments, the same or similar elements or steps will use the same or similar labels.

Please refer to FIG. 1, which is a schematic diagram of the steps of the first embodiment of the heart evaluation method of the present invention. The heart evaluation method of the present embodiment includes a database establishment phase S100 and an evaluation phase S200, wherein the database establishment phase S100 can establish a mapping tensor and a database as a basis for calculation in the evaluation phase, and includes the following steps: establishing a model of a virtual heart model (reference heart) (step S110); simulating different degrees of abnormality at different locations of the model and generating a plurality of sample signals (step S120); classifying the sample signals, generating a plurality of corresponding modal vectors, and establishing a mapping tensor according to the modal vectors (step S130). On the other hand, the evaluation stage S200 can receive the ECG signal of the subject, analyze and classify the ECG signal, and includes the following steps: providing an ECG signal (step S210); obtaining a body surface potential vector according to the ECG signal (step S220); calculating the body surface potential vector and the mapping tensor to obtain a heart potential vector (step S230); and classifying the abnormal position and degree of abnormality of an evaluated heart corresponding to the ECG signal according to the heart potential vector (step S240).

In detail, the human body's cardiac potential signals are transmitted to the body surface and can be detected by the electrical signal electrodes of non-invasive devices to obtain body surface signals. In other words, there is a specific mapping relationship between cardiac potential signals and body surface signals. The electrocardiogram detection method is to read the body surface signals obtained by the instrument, and professional medical personnel can combine past professional knowledge and detection experience to interpret and confirm the heart status of the subject. The cardiac assessment method of this embodiment adopts traditional electrophysiological theory, that is, a virtual heart model is first established, and the virtual heart model simulates the heart state to be assessed to generate a corresponding cardiac potential signal. The generated cardiac potential signal is then converted into a body surface signal through a mapping tensor, and the model and the mapping tensor are adjusted and corrected so that the generated body surface signal is close to the actual situation. Once the adjustment and correction of the virtual heart model is completed, there is no need to connect the instrument to the human body in real time for future assessments. It is only necessary to obtain the body surface signals measured from the subject. Through the established model and mapping tensor, it can be known whether the subject's body surface signals are consistent with the body surface signals generated when the virtual heart model is abnormal, thereby confirming whether the subject's heart is abnormal, the location of the abnormality, and even the degree of the abnormality.

Therefore, in the database establishment stage S100, a three-dimensional model of the virtual heart is first established. This embodiment obtains discrete cross-sectional images of a real heart in a specific direction through computer tomography, and simulates the topological shape and internal structure of the real heart through filtering (such as median filtering), adjusting imaging parameters (such as Hounsfield units) and smoothing, combined with the finite element method (FEM), i.e. the above-mentioned virtual heart model.

Afterwards, the time series of cardiac potential excitation is reconstructed based on the electrophysiological theory of the heart. Since the real heart includes different nerve nodes, nerve fibers and muscle fibers, multiple time series can be obtained by simulating the conductivity rate and conductivity pathway of these tissues. The actual pathway of each current flow is calculated by Poisson’s equation, and different potential time series of each tissue are obtained. Sample signals are further generated based on the electrophysiological theory of the real heart.

Please refer to Figure 2, which is a schematic diagram of (a) the left anterior descending artery; (b) the left circumflex artery; (c) the anterior branch of the right coronary artery; and (d) the posterior branch of the right coronary artery in a virtual heart model. When the abnormal symptom of interest is myocardial infarction, one of the causes of myocardial infarction is the sclerosis or embolism of these arteries that supply oxygen and blood to the myocardium, resulting in myocardial ischemia and failure to contract normally. Therefore, in this embodiment, the virtual heart model is initially classified into the left anterior descending (LAD) artery, the left circumflex (LCX) artery, and the right coronary artery (RCA), and each category is further divided into three, three, and five regions, so as to simulate the potential state of ischemia in different regions. Based on clinical experience and the upstream and downstream relationship of blood flow in the artery, 26 ischemic regional patterns (abnormal locations) can be defined as shown in Table 1.

Specifically, the cardiac evaluation method of the present embodiment mainly uses 12-lead electrocardiogram signals as input parameters for evaluation, but the present invention is not limited thereto. According to actual hardware specifications or more accurate requirements for evaluation results, the electrocardiogram signals used in the cardiac evaluation method can also use electrocardiogram signals with lower lead numbers (e.g., single lead to 11 leads), which can effectively reduce the required time and cost; or use electrocardiogram signals with more leads (e.g., 24, 48 or 64 leads or more) to obtain more accurate results through high-performance computers. In addition, the evaluated diseases are not limited to myocardial infarction, but can also include heart failure, ventricular hypertrophy, etc. Further considering the actual distribution of blood vessels at the borders of adjacent regions, the above 26 ischemic regional patterns can be further expanded or reduced, and random pairing can be used to increase the possibility of regional combinations, thereby generating 1638 non-repetitive 12-lead potential distributions. These 1638 non-repetitive potential distributions are then combined with 5 ischemic severity levels (abnormality levels) to further combine 8190 12-lead ECG signal databases, i.e., the sample signals shown in step S120 in Figure 1.

It is worth mentioning that the severity of the above five types of ischemia is determined by transcellular potential. In this embodiment, the reduction of resting potential amplitude (for example, to 93%, 90%, 85%, 80% and 50% of the normal potential amplitude) and the delay of the excitation potential time are used as the index values of severity 1-5, but the present invention is not limited to this. According to the actual needs of the assessor or the assessment of symptoms, the user can also define different severity classifications by different index values as the basis for establishing a database.

Since the surface ECG signal is a quasi-periodic signal, the simulated ischemic surface ECG signal can be normalized and arranged through machine learning (such as ANN neural network) or artificial methods, so that the 8190 ECG signals corresponding to multiple abnormal locations and multiple abnormal degrees of the virtual heart model (reference heart) form a tensor. Since each abnormal location and each combination of abnormal degrees corresponds to a mode of the virtual heart, the vector corresponding to these ECG signals is defined as the mode vector v . In this way, the mode vector v can be arranged into the above-mentioned mapping tensor A according to the corresponding artery, that is:

Where A _Anterior , A _Lateral and A _Inferior represent the sub-mapping tensors of the anterior artery, lateral artery and inferior artery in the mapping tensor A, respectively, and v _A , v _L and v _I represent the weight vectors in each sub-mapping tensor. Therefore, if the body surface potential is represented in vector form, that is, a body surface potential vector y is defined and the body surface potential vector y corresponds to a heart potential vector x of the virtual heart model, then the mapping tensor A, the heart potential vector x and the body surface potential vector y should satisfy the following expression, namely: Ax = y

Since the mapping tensor A and the body surface potential vector y are known, it is only necessary to solve for the heart potential vector x , and then the mode to which the evaluated heart corresponds to the provided complex lead electrocardiogram signal can be obtained, and the abnormal location and degree of the abnormality of the evaluated heart can be classified, or the above classification results can be provided to medical personnel as a reference for judging myocardial infarction.

Please refer to FIG. 1 and FIG. 3 , wherein FIG. 3 is a schematic diagram of the step flow of the second embodiment of the heart assessment method of the present invention. The heart assessment method of this embodiment is roughly similar to the heart assessment method of the first embodiment, and the main difference is that the database establishment stage S100a of the heart assessment method of this embodiment also includes providing a plurality of real case signals, and verifying the mapping tensor through the real case signals (step S140).

In detail, during the establishment of the mapping tensor A and the database, the electrocardiogram signals corresponding to the real cases of myocardial infarction can be obtained from the clinical cases, and these electrocardiogram signals are solved through the mapping tensor A to obtain the corresponding cardiac potential signals, and it is confirmed whether the cardiac mode corresponding to the obtained cardiac potential signal is consistent with the abnormal position and degree of abnormality on the virtual heart model, thereby verifying the mapping tensor A and further improving the reliability of the calculation model. In this embodiment, the number of real case signals used for verification with the mapping tensor A is 148, but according to actual needs and the number of cases obtained, the real case signals used for verification can be increased or decreased, and the present invention does not limit this.

Please refer to Figures 3 and 4, wherein Figure 4 is a schematic diagram of the step flow of the third embodiment of the heart evaluation method of the present invention. The heart evaluation method of this embodiment is roughly similar to the heart evaluation method of the second embodiment, and the main difference is that: the evaluation stage S200a of the heart evaluation method of this embodiment also includes pre-processing the electrocardiogram signal to obtain a plurality of local potential signals (step S220a); and extracting at least one characteristic value from each of these local potential signals, and obtaining the body surface potential vector based on these characteristic values (step S220b).

Please refer to Figures 4 and 5, where Figure 5 is a schematic diagram of extracting feature values in step S220b of Figure 4. Since the original electrocardiogram signal measured from the subject may contain noise, and the feature values required for its calculation are not obvious. Therefore, the original electrocardiogram signal can be pre-processed to improve the accuracy of the subsequent calculation results.

In detail, the electrocardiogram signal used in this embodiment is, for example, a 12-lead surface electrocardiogram signal. These electrocardiogram signals can be directly measured from the subject in real time, or can be measured in advance and stored in a computer or hospital database and obtained through transmission and reception. The present invention is not limited to this. When the subject's original electrocardiogram signal has noise, a bandwidth filter of, for example, 0.05-40 Hz can be used to filter out arrhythmias and unaligned heartbeat signals, and further a median filter is used to discretely sample the original signal to obtain a cleaner signal. Then, the QRS wave in the signal is detected by the instrument, making the quasi-periodic characteristics of the signal more obvious. In addition, the Pan Tompkin algorithm can be used to determine the time interval between adjacent cycles, thereby extracting a single cycle signal from the quasi-periodic signal. Finally, the oscillation in the periodic signal can be averaged and smoothed to obtain a local potential signal that corresponds to the original electrocardiogram signal but is clearer.

After completing the signal pre-processing step, the eigenvalues required for calculation can be further extracted from these local potential signals. Specifically, since the ST segment in the electrocardiogram signal has a strong correlation with the manifestation of myocardial infarction, the present embodiment extracts local potential signals from the electrocardiogram signal, and extracts 5 eigenvalues for each ST segment of these local potential signals. These eigenvalues include, but are not limited to, the potential amplitude of the J point _P1 , the minimum amplitude of the lowest point of the T wave _P2 , the maximum amplitude of the highest point of the T wave _P3 , the potential amplitude of the first potential point _P4 , and the potential amplitude of the second potential point _P5 , wherein the first potential point _P4 is a potential point on the ST segment of this local potential signal that is 0.25 times the JT interval away from the J point _P1 , and the second potential point _P5 is a potential point on the ST segment of this local potential signal that is 0.5 times the JT interval away from the J point _P1 . Because each lead has these five eigenvalues, 60 eigenvalues can be obtained from the above 12-lead ECG signal and arranged in sequence in vector form to obtain a 60×1 body surface potential vector y .

R ^60×8190,x

R ^8190×1,y

R ^60×1 Therefore, in the heart evaluation method of the third embodiment, the condition of the equation can be further established as: Ax = y A

R ^60×8190 , x

R ^8190×1 , y

R ^60×1

Please refer to FIG. 4 and FIG. 6 , where FIG. 6 is a schematic diagram of the step flow of the fourth embodiment of the heart assessment method of the present invention. The heart assessment method of this embodiment is roughly similar to the heart assessment method of the third embodiment, and the main difference is that: the assessment stage S200b of the heart assessment method of this embodiment also includes determining at least one modal vector corresponding to the body surface potential vector among these modal vectors through machine learning (step S230a); obtaining at least one weighted eigenvalue of the body surface potential vector on the modal vector and at least one corresponding weight vector (step S230b); and combining the weighted eigenvalue and the weight vector to obtain the heart potential vector (step S230c).

可被分類為第i類 的模態，其中i

[1,130]，且以上述前支動脈、側支動脈以及下支動脈的順序排列各個模態向量v時，可進一步得到以下方程式：

In detail, in the cardiac assessment method of the third embodiment, it can be found that there are only 60 eigenvalues used as the basis for assessment, but even the ischemic location and ischemic degree combinations that have not been expanded or reduced by nodes have 26×5=130 combinations. Therefore, if this indeterminate equation is solved directly, it may not be possible to obtain a unique solution, or a large number of boundary conditions need to be set to solve it, and a large amount of computing resources and time are required to solve it. However, according to the traditional electrophysiological simulation process, it can be known that a specific arterial embolism has no obvious interaction with other arteries. Therefore, if it is assumed that the surface potential vector y

can be classified as the i- th mode, where i

[1,130], and when the modal vectors v are arranged in the order of the anterior artery, the lateral artery, and the inferior artery, the following equation can be further obtained:

Wherein ω _i and vi are respectively at least one weighted eigenvalue and at least one corresponding weight vector on at least one modal vector v corresponding to the body surface potential vector y _in the modal vector v , and the corresponding weight vector can be determined by machine learning, such as sparse representation classification, so as to reduce the 8190 modal vectors v that the body surface potential vector y may correspond to to a smaller number of weight vectors vi _, thereby reducing the unknown conditions when solving. In addition, _the heart potential vector x0 is defined as:

Then the mapping tensor A, the heart potential _vector x0 and the body surface potential vector y can be expressed as follows:

Because the dimension of the range 60 is much smaller than the row space dimension 8190 of the mapping tensor A, and most of the elements in the mapping tensor A are 0, x0 _is sparse enough to conform to the _Lagrangian form of the Lasso algorithm. Therefore, the loss function L can be further defined by the coordinate descent method, that is: x ⁼ [ x1 , x2 _, …, x8190 ] _T .

In order to obtain the minimum value of the loss function L , the loss function L can be partially differentiated with respect to any element x _L in the cardiac potential vector x ₀ , and the element x _L = xi obtained is the weighted eigenvalue ω _i of the corresponding category, so the body surface potential vector y _can be classified into a linear combination of _a specific weighted eigenvalue ω _i and a weight vector vi . Because each mode corresponds to a different distribution of ischemic locations and different degrees of ischemia, users can use less computing resources to make a highly accurate classification of the ischemic location and degree of ischemia in a manner close to analytical solution, greatly improving the performance of the cardiac assessment method.

Please refer to FIG. 6 and FIG. 7 , where FIG. 7 is a schematic diagram of the step flow of the fifth embodiment of the heart assessment method of the present invention. The heart assessment method of this embodiment is roughly similar to the heart assessment method of the fourth embodiment, and the main difference is that the heart assessment method of this embodiment also includes establishing a three-dimensional virtual heart image of the virtual heart according to the abnormal position and abnormal degree in the assessment stage S200c (step S250).

Please refer to FIG. 7 and FIG. 8 , where FIG. 8 shows (a) a topography diagram of the three-dimensional virtual heart image created in step S250 of FIG. 7 ; and (b) a colored topography diagram. When the user obtains the category of the modal vector v corresponding to the assessed heart through the cardiac potential vector x , the ischemic location and degree of the assessed heart can be known. At this time, the modal vector v corresponding to the cardiac potential vector x in the mapping tensor A can be combined with the virtual heart model and projected into the three-dimensional coordinate space to form a topological map established by finite elements, or different cardiac regions in the topological map can be given different colors according to the degree of ischemia, which can provide medical personnel with visual information similar to cardiac nuclear medicine examination angiography, which is more conducive to medical personnel's judgment.

Please refer to Figures 7 and 9, wherein Figure 9 is a schematic diagram of the step flow of the sixth embodiment of the heart assessment method of the present invention. The heart assessment method of this embodiment is roughly similar to the heart assessment method of the fifth embodiment, and the main difference is that the heart assessment method of this embodiment also includes projecting the three-dimensional virtual heart image onto a plane and filtering it in the assessment stage S200d to obtain a two-dimensional heart zoning image (step S260).

r=z _n /D x _n '=x _n ×r y _n '=y _n ×r n

[1,2,…,257] Please refer to FIG. 9 and FIG. 10 , where FIG. 10 is a schematic diagram of the two-dimensional heart zoning images of (a) the first subject; (b) the second subject; and (c) the third subject obtained in step S260 of FIG. 9 . Although the three-dimensional virtual heart image can fully present the ischemic area and degree of ischemia of the heart, medical personnel need to look around the entire virtual heart model before confirming the complete condition of the heart, and cannot quickly and intuitively present medical diagnosis information. Therefore, the three-dimensional virtual heart image can be further projected into a two- _dimensional image. Taking the circular image as an example, the coordinates ₍ xn _, yn , zn ) of the three-dimensional virtual heart image can be determined by the following formula to correspond to the _new coordinates ( xn _' , yn ') in the two-dimensional heart image:

r = z _n / D x _n '= x _n × ry _n '= y _n × rn

[1,2,…,257]

After obtaining the new coordinates ( xn _' , yn _' ), the rotation angle between the new coordinates ( xn _' , yn _' ) and the original coordinates ( xn _, yn ) can be determined by the rotation matrix, that _is :

After the projection is completed, it can be processed by graphic masking, filtering (such as a median filter) and color adjustment to form a two-dimensional heart zoning image as shown in Figure 10 (a), Figure 10 (b) and Figure 10 (c). The different blocks in the circular graph from the top along the clockwise direction represent the right posterolateral artery, lateral artery, anterior artery and inferior artery of the virtual heart model. Dark colors represent sufficient blood flow, while lighter colors represent more severe ischemia. In this way, medical personnel can quickly confirm and evaluate the ischemic location and degree of the heart, further improving the efficiency of the heart assessment method.

It is worth mentioning that, although in this embodiment, a three-dimensional virtual heart image is first created and then projected onto a plane to form a two-dimensional heart zoning image, the present invention is not limited thereto. When the user does not need to refer to a three-dimensional virtual heart image, the abnormal position and degree of abnormality corresponding to the heart potential _vector x0 can also be directly mapped onto a plane to form an image, thereby saving the time and resources of constructing a three-dimensional model.

In fact, the heart assessment method provided by the present invention can be implemented by a computer program, or a computer-readable recording medium (such as a CD, USB or hard disk, etc.) storing the above computer program. When the computer loads the above computer program and executes it, the following steps can be executed: providing an electrocardiogram signal; obtaining a body surface potential vector based on the electrocardiogram signal; calculating the body surface potential vector with a pre-established mapping tensor to obtain a heart potential vector, wherein the mapping tensor is obtained based on a reference heart to generate a plurality of abnormal locations and a plurality of abnormal degrees; and classifying the abnormal location and abnormal degree of an assessed heart corresponding to the electrocardiogram signal based on the heart potential vector.

Specifically, the computer program may include a receiving unit, an evaluation unit, a database, and an output unit, wherein the receiving unit is used to receive the measured or pre-stored electrocardiogram signal; the evaluation unit may calculate the electrocardiogram signal received by the receiving unit and the mapping tensor A established according to the modal vector v in the database through the above-mentioned heart evaluation method to obtain the heart potential vector x , and after classifying the abnormal position and degree of abnormality of the evaluated heart, the classification result is output through the output unit, or further displayed as shown in Figures 8 and 10. Since the above process is realized through program automation, the manpower and cost required for general detection are greatly reduced, so it can be applied to early prediction and treatment tracking.

The present invention has been disclosed in the above text with preferred embodiments, but those familiar with the present technology should understand that the above embodiments are only used to describe the present invention and should not be interpreted as limiting the scope of the present invention. It should also be noted that all changes and substitutions equivalent to the above embodiments should be included in the scope of the present invention. Therefore, the scope of protection of the present invention shall be based on the scope defined by the patent application.

S100: Database creation phase

S200: Evaluation phase

S110~S130, S210~S240: Steps

Claims (7)

A heart assessment method includes the following steps: providing an electrocardiogram signal; obtaining a body surface potential vector according to the electrocardiogram signal; calculating the body surface potential vector with a pre-established mapping tensor, wherein the mapping tensor simulates a plurality of sample signals generated by different degrees of ischemia at different locations of a model of a virtual heart, classifying the plurality of sample signals to generate corresponding plurality of modal vectors, and establishing a method based on the plurality of modal vectors; using machine learning and sparse representation classification to determine at least one modal vector corresponding to the body surface potential vector; using coordinate descent method to obtain a modal vector corresponding to the body surface potential vector; Descent) to obtain at least one weighted eigenvalue of the body surface potential vector on the at least one modal vector and at least one corresponding weight vector; and classify the ischemic position and degree of ischemia of an assessed heart corresponding to the electrocardiogram signal according to the cardiac potential vector. The heart assessment method as described in claim 1 further includes the following steps: pre-processing the electrocardiogram signal to obtain a plurality of local potential signals, wherein the number of the plurality of local potential signals is the same as the number of leads of the electrocardiogram signal; and extracting at least one characteristic value from each of the plurality of local potential signals, and obtaining the body surface potential vector based on the plurality of at least one characteristic value. A heart assessment method as described in claim 2, wherein the at least one characteristic value includes a J point amplitude, a T wave minimum amplitude, a T wave maximum amplitude, a first potential point amplitude, and a second potential point amplitude of a local potential signal corresponding to the plurality of local potential signals, wherein the first potential point is a potential point at 0.25 times the J-T interval on the S-T segment of the local potential signal, and the second potential point is a potential point at 0.5 times the J-T interval on the S-T segment of the local potential signal. The heart assessment method as described in claim 1, wherein the reference heart is a virtual heart, and the heart assessment method further comprises the following steps: establishing a three-dimensional virtual heart image of the virtual heart according to the ischemic location and the ischemic degree. The heart assessment method as described in claim 4 further includes the following steps: projecting the three-dimensional virtual heart image onto a plane and filtering it to obtain a two-dimensional heart zoning image. A cardiac assessment method as described in claim 1, wherein the number of leads of the electrocardiogram signal is less than 13. A computer program product, when loaded into a computer and executed, can perform the steps performed by the cardiac assessment method as described in any one of claims 1-6.